Antibiotics for acute bronchitis
Abstract
Background
The benefits and risks of antibiotics for acute bronchitis remain unclear despite it being one of the most common illnesses seen in primary care.
Objectives
To assess the effects of antibiotics in improving outcomes and assess adverse effects of antibiotic therapy for patients with a clinical diagnosis of acute bronchitis.
Search methods
We searched CENTRAL 2013, Issue 12, MEDLINE (1966 to January week 1, 2014), EMBASE (1974 to January 2014) and LILACS (1982 to January 2014).
Selection criteria
Randomised controlled trials (RCTs) comparing any antibiotic therapy with placebo or no treatment in acute bronchitis or acute productive cough, in patients without underlying pulmonary disease.
Data collection and analysis
At least two review authors extracted data and assessed trial quality.
Main results
Seventeen trials with 5099 participants were included in the primary analysis. The quality of trials was generally good. There was limited evidence to support the use of antibiotics in acute bronchitis. At follow‐up, there was no difference in participants described as being clinically improved between antibiotic and placebo groups (11 studies with 3841 participants, risk ratio (RR) 1.07, 95% confidence interval (CI) 0.99 to 1.15; number needed to treat for an additional beneficial outcome (NNTB) 22. Participants given antibiotics were less likely to have a cough (four studies with 275 participants, RR 0.64, 95% CI 0.49 to 0.85; NNTB 6); have a night cough (four studies with 538 participants, RR 0.67, 95% CI 0.54 to 0.83; NNTB 7) and a shorter mean cough duration (seven studies with 2776 participants, mean difference (MD) ‐0.46 days, 95% CI ‐0.87 to ‐0.04). The differences in presence of a productive cough at follow‐up and MD of productive cough did not reach statistical significance.
Antibiotic‐treated patients were more likely to be improved according to clinician's global assessment (six studies with 891 participants, RR 0.61, 95% CI 0.48 to 0.79; NNTB 25); were less likely to have an abnormal lung exam (five studies with 613 participants, RR 0.54, 95% CI 0.41 to 0.70; NNTB 6); have a reduction in days feeling ill (five studies with 809 participants, MD ‐0.64 days, 95% CI ‐1.16 to ‐0.13) and a reduction in days with limited activity (six studies with 767 participants MD ‐0.49 days, 95% CI ‐0.94 to ‐0.04). The differences in proportions with activity limitations at follow‐up did not reach statistical significance. There was a significant trend towards an increase in adverse effects in the antibiotic group (12 studies with 3496 participants) (RR 1.20, 95% CI 1.05 to 1.36; NNT for an additional adverse effect 24).
Authors' conclusions
There is limited evidence to support the use of antibiotics in acute bronchitis. Antibiotics may have a modest beneficial effect in some patients such as frail, elderly people with multimorbidity who may not have been included in trials to date. However, the magnitude of this benefit needs to be considered in the broader context of potential side effects, medicalisation for a self‐limiting condition, increased resistance to respiratory pathogens and cost of antibiotic treatment.
PICOs
Plain language summary
Antibiotic treatment for people with a clinical diagnosis of acute bronchitis
Acute bronchitis is a clinical diagnosis for an acute cough, which may or may not be productive of mucus or sputum. It occurs when the tubes (bronchi) within the lungs become inflamed and may be caused by viruses or bacteria. Symptoms generally last for two weeks but the associated cough can last for up to eight weeks. Recently, there has been controversy over the term acute bronchitis as it covers a range of clinical presentations that may overlap with other diagnoses such as upper or lower respiratory tract infections. For this reason, some have suggested using the term 'acute lower respiratory tract infection when pneumonia is not suspected' as this is more specific. Antibiotics are commonly prescribed to treat this condition though other treatments providing symptom relief are commonly used. Antibiotics can have adverse effects such as nausea and diarrhea but can cause more serious reactions related to anaphylaxis in those allergic to them. In healthy communities, there is little evidence of bacterial infection in people with bronchitis and there is no practical test to distinguish between bacterial and viral bronchitis. Within this context the use of antibiotics to treat acute bronchitis is controversial but common. Concerns that prescribing unnecessary antibiotics increases antibiotic resistance exists.
We included 17 trials with 3936 participants diagnosed with acute bronchitis and randomly assigned to receive any antibiotic treatment or a placebo or no treatment. Co‐treatments with other medications to relieve symptoms were allowed if they were given to all patients. We excluded patients with pre‐existing underlying pulmonary disease such as chronic bronchitis or chronic obstructive pulmonary disease. The quality of trials was generally good, particularly for more recent studies. There was limited evidence to support the use of antibiotics for acute bronchitis and a large study involving 1038 patients from 12 countries included in this update has confirmed this finding. Some people treated with antibiotics recovered a bit more quickly with reductions in cough‐related outcomes though the difference was of doubtful clinical significance as it amounted to a difference of half a day over an 8 to 10 day period. There was a statistically significant but small increase in adverse side effects in patients treated with antibiotics. The most commonly reported side effects included nausea, vomiting or diarrhea, headaches, skin rash and vaginitis.The available evidence suggests that there is no benefit in using antibiotics for acute bronchitis in otherwise healthy individuals though more research is needed on the effect in frail, elderly people with multimorbidities who may not have been included in the existing trials. The use of antibiotics needs to be considered in the context of the potential side effects, medicalisation for a self‐limiting condition and costs of antibiotic use, particularly the potential harms at population level associated with increasing antibiotic resistance.
Authors' conclusions
Background
Description of the condition
Acute bronchitis is a common illness which is characterised by fever and cough that is often wheezy in nature and the cough may or may not be productive. Acute bronchitis occurs when the bronchi become inflamed and it may be caused by either viral or bacterial infection. Symptoms generally last for two weeks but the associated cough can last for up to eight weeks (CDC 2013). It is the ninth most common among outpatient illnesses recorded by physicians in ambulatory practice in the USA (Delozier 1989) and the fifth most commonly encountered by Australian General Practitioners, for whom it represents 3.5% of encounters and 2.4% of problems seen (Meza 1994). In the UK, there are 300 to 400 consultations for treatment of respiratory tract infections per 1000 registered patients each year and while antibiotic prescribing for these conditions had declined between 1995 and 2000, it has since stabilised (Gulliford 2011). The European Centre for Disease Prevention and Control provides data on trends in antimicrobial consumption across Europe suggesting that overall antibiotic use varies across Europe with most countries showing increases over the period 1997 to 2010 (ECDC 2013).
Population‐based estimates of the incidence of acute bronchitis range from 33 to 45 cases per 1000 per year (Ayres 1986; Mainous 1996). Patients with bronchitis miss an average of two to three days off work per episode. The great majority of episodes of acute bronchitis in healthy individuals are presumed to be viral infections, although this has been questioned (Macfarlane 1994). Community‐based studies have isolated viruses in 8% to 23% of cases (Boldy 1990; Macfarlane 1993; Stuart‐Harris 1965). Other pathogens implicated in acute bronchitis are Mycoplasma pneumoniae (M. pneumoniae), Chlamydia pneumoniae (C. pneumoniae) and Bordetella pertussis (B. pertussis), each of which has been identified in up to 25% of cases in various populations (Boldy 1990; Falck 1994; Foy 1993; Grayston 1993; Herwaldt 1991; Jonsson 1997; King 1996; Macfarlane 1993; Robertson 1987; Stuart‐Harris 1965; Thom 1994). A more recent study assessing the aetiology and outcome of acute lower respiratory tract infection in 638 adults in UK primary care, showed that in 55% viral or bacterial pathogens were identified (Macfarlane 2001).
Description of the intervention
The use of antibiotics for patients with acute bronchial infections remains a controversial area in primary health care practice (Coenen 2007; Gonzales 1995). Streptococcus pneumoniae (S. pneumoniae), Haemophilus influenzae (H. influenza) andMoraxella catarrhalis (M. catarrhalis) have been isolated from sputum samples in up to 45% of patients with acute bronchitis (Henry 1995; Macfarlane 1993) but their role is difficult to assess because of potential oropharyngeal colonisation in healthy individuals (Laurenzi 1961; Smith 1986). Unfortunately, there are no clinically useful criteria that accurately help distinguish bacterial from viral bronchial infections. Therefore, some authors have called for physicians to stop prescribing antibiotics for patients with acute bronchitis (Gonzales 1995; Hueston 1997). Nonetheless, antibiotics are prescribed for 60% to 83% of patients who present to physicians with this disorder (Gonzales 1997; Mainous 1996; Meza 1994; Petersen 2007; Straand 1997). Overall antibiotic use varies across Europe with most countries showing increases over the period 1997 to 2010 (ECDC 2013).
How the intervention might work
Antibiotics may improve outcomes in acute bronchitis if the disease is caused by a bacterial infection. They have no antiviral activity so are not effective in viral bronchitis. In addition, antibiotics can cause harm relating to their adverse effect on normal bacteria colonising the intestine. These adverse effects most commonly include gastrointestinal symptoms such as nausea and diarrhea but antibiotics can also cause more serious reactions related to anaphylaxis in those allergic to them.
Why it is important to do this review
As acute bronchitis occurs so frequently, it is important to obtain some estimate of the probable effectiveness of antibiotic therapy. If effective, antibiotics could shorten the course of the disease and reduce the loss of productive work time it causes. However, any benefit from antibiotics must be weighed against the possibility that excessive antibiotic use will lead to increases in cost and patient morbidity, as well as development of resistant strains of common organisms (Coenen 2007; Molstad 1992) and unnecessary medicalisation of individuals with a self‐limiting illness (Little 2005). If antibiotics are ineffective, then their use should be discontinued.
Objectives
To assess the effects of antibiotics in improving outcomes and assess adverse effects of antibiotic therapy for patients with a clinical diagnosis of acute bronchitis.
Methods
Criteria for considering studies for this review
Types of studies
Randomised controlled trials (RCTs) in patients with acute bronchitis assigned to treatment with an antibiotic or a placebo or no active treatment.
Types of participants
We included trials that included patients of either sex or any age with a clinical syndrome of cough with or without productive sputum, with a physician's diagnosis of acute bronchitis or cough with persistent cold or flu‐like illness that was not resolving. The term "acute lower respiratory tract infection when pneumonia is not suspected" is also used to describe this clinical presentation. We excluded trials that included patients with pre‐existing chronic bronchitis (i.e. acute exacerbation of chronic bronchitis).
Types of interventions
We included all RCTs comparing any antibiotic therapy versus no treatment or placebo in the management of acute bronchitis. We excluded trials comparing one antibiotic regimen with another, or trials comparing the use of other active medications (such as bronchodilators) with antibiotic therapy in this review. We included trials that allowed concurrent use of other medications such as analgesics, antitussives, antipyretics or mucolytics if they allowed equal access to such medications for patients in the antibiotic and control groups.
Types of outcome measures
We included the following range of cough‐related and general clinical outcomes.
Primary outcomes
-
Cough‐related outcomes including:
-
time to resolution of cough;
-
sputum production, defined as proportion of patients with or without sputum;
-
proportions of patients with cough, night cough, productive cough.
-
-
Global assessment of improvement by clinicians at follow‐up.
-
General clinical outcomes including:
-
severity of symptoms;
-
activity limitations;
-
abnormal lung examination at a designated follow‐up visit.
-
Secondary outcomes
-
Adverse effects.
Search methods for identification of studies
Electronic searches
For this updated review, we searched the Cochrane Central Register of Controlled Trials (CENTRAL) 2013, Issue 12, part of The Cochrane Library, www.thecochranelibrary.com (accessed 15 January 2014), which includes the Acute Respiratory Infections (ARI) Group's Specialised Register, MEDLINE (1966 to January week 1, 2014), EMBASE (1974 to January 2014) and LILACS (1982 to January 2014). We used the search strategy described in Appendix 1 to search MEDLINE and CENTRAL. The search strategy was adapted to search EMBASE (Appendix 2) and LILACS (Appendix 3). Details of the 2007 update search are in Appendix 4.
Searching other resources
We searched the WHO International Clinical Trials Registry Platform and ClinicalTrials.gov (19 February 2013). We also searched the reference lists of relevant trials, review articles and textbook chapters to identify additional trials, including those published prior to 1966; and we included articles from the authors' personal collections. We also requested unpublished trials from trial authors. In addition, we also contacted drug companies that manufacture antibiotics for the earlier version of this review. There were no language or publication restrictions.
Data collection and analysis
Selection of studies
For the original review (Becker 1997), two authors (LB and JS) independently used the titles and abstracts of the identified citations to exclude trials that clearly did not meet the inclusion criteria of the review. The full paper was obtained for further examination if either review author felt that the trial might possibly meet the criteria. The most common reasons for exclusion were the lack of a control group receiving a placebo, or the inclusion of patients with chronic bronchitis.
Three review authors (LB, JS, TF) reviewed articles, which passed this initial screen by using only the method section of each paper, without reference to the names of the authors, the institution, the journal or the results, to determine their fit with the inclusion criteria for this review. Nine articles passed this second screen but only eight had extractable data. We obtained unpublished data from the other trial (Hueston 1994) from the trial author.
For review updates, two authors screened citations after an updated search was carried out by the Acute Respiratory Infections (ARI) Group for the 2004 update of this review (Fahey 2004). We incorporated two RCTs that met the inclusion criteria into the review. An updated search was carried out by the ARI Group for the 2009 update of the review (Smith 2009). Two review authors (TF, SS) identified and screened three potentially eligible papers and one additional study was included (Little 2005). The ARI Group updated the searches for the Smith 2011 update and one additional eligible study was included (Nduba 2008).
For this 2014 updated review, two review authors screened and identified one new eligible study (Little 2013) and one ongoing study (Stocks 2013). One of the ongoing trials has been published and is also included in the review (Llor 2013).
Data extraction and management
More than one review author independently extracted data from original studies for each update of this review. We resolved disagreements by discussion between the review authors.
Assessment of risk of bias in included studies
Three review authors (LB, KJS, TF) evaluated the methodological quality of each trial, while remaining blinded to the names of the trial authors, the institution and the journal in which the trial was published. For previous versions of the review, we assessed the methodological quality of each trial using a scoring system described by Chalmers (Chalmers 1990), which assigned points for randomisation method, blinding of outcome assessment and of patients, intention‐to‐treat (ITT) analysis, contamination and co‐intervention and losses to follow‐up. Agreement among the review authors regarding the quality of the articles was high. Disagreements were resolved by discussion and consensus.
For the 2009 update, two review authors (SS, TF) reassessed the original quality of all included studies by using the new 'Risk of bias' guidelines and incorporated this into the 'Risk of bias' tables now presented in this updated review. Two review authors (SS, TF) screened the studies added in the 2011 and 2013 updates.
Measures of treatment effect
The effect measures of choice were risk ratio (RR) for categorical outcomes and mean difference (MD) for continuous data.
Unit of analysis issues
There were no cluster‐randomised trials included in this review as it involved a simple drug trial with a placebo comparator. Clinicians were generally blinded to the intervention. We identified no unit of analysis errors.
Dealing with missing data
Where data were missing this was reported within the risk of bias section. We did not adopt any strategies to deal with missing data such as imputation. In general, missing data did not bias the review findings.
Assessment of heterogeneity
Where clinical heterogeneity was considered to be an issue, we undertook a random‐effects meta‐analysis rather than a fixed‐effect meta‐analysis. This particularly applied to the most recent analysis added to this updated version of the review (Analysis 6.1).
Assessment of reporting biases
We examined funnel plots for each of the included analyses and none indicated any significant level of reporting bias.
Data synthesis
All previous versions of this review have presented fixed‐effect meta‐analyses. For this update, we included a range of outcomes under the broad definition of 'clinically improved'. These were clinically heterogeneous so we used a random‐effects meta‐analysis.
Subgroup analysis and investigation of heterogeneity
We also carried out a subgroup analysis comparing studies using a placebo control or no active treatment.
Sensitivity analysis
We included only studies that limited enrolment to patients with a clinical diagnosis of acute bronchitis or acute productive cough for the primary analysis. We did a sensitivity analysis that included unpublished data from subgroups of patients with a productive cough (Howie 1970) and non‐purulent tracheobronchitis (Kaiser 1996) from two studies that enrolled patients with an influenza‐like illness or a common cold.
Results
Description of studies
Results of the search
The updated and modified CENTRAL, MEDLINE, EMBASE and LILACS searches in 2014 yielded an additional 799 titles. All of the 17 trials included in the primary analysis enrolled patients with a diagnosis of acute cough or acute lower respiratory tract infection. In one study (Franks 1984), patients were required to produce a sputum sample for analysis as a condition of enrolment.
Included studies
For this 2014 update, two new studies were added (Little 2013; Llor 2013). These were important additions, particularly the trial by Little 2013 as it is the largest trial conducted to date and included 2061 patients recruited across 12 countries. Patients were randomised to receive amoxicillin or placebo and there was low risk of bias with more than 80% follow‐up of participants.
For most studies, clinical findings were used to exclude patients thought to have pneumonia. Four studies included chest radiographs in their protocols ‐ two (Brickfield 1986; Nduba 2008) performed a chest film on all potential participants. In the other two, Scherl 1987 did so on patients with rales or fever and Llor 2013 did so on cases with suspected pneumonia (seven of 416 participants). Both excluded those with radiological evidence of pneumonia or tuberculosis (TB). One study (Stott 1976) excluded patients with any abnormality noted on examination of the chest. Four trials also excluded patients with a clinical syndrome suggesting sinusitis (Dunlay 1987; King 1996; Verheij 1994; Williamson 1984).
In all trials, the duration of illness at entry was less than 30 days. One trial (Stott 1976) limited enrolment to patients ill for less than one week; in five trials the duration was two weeks or less (Brickfield 1986; Evans 2002; Franks 1984; King 1996; Matthys 2000).
Eight of the trials included only adults (Brickfield 1986; Dunlay 1987; Hueston 1994; Little 2013; Llor 2013; Nduba 2008; Verheij 1994; Williamson 1984). The remaining studies included adolescents plus adults (Franks 1984; Scherl 1987; Stott 1976) or patients aged three years (Little 2005) or eight years or older (King 1996).
As for antibiotic treatment, four trials (Scherl 1987; Stott 1976; Verheij 1994; Williamson 1984) used doxycycline, four erythromycin (Brickfield 1986; Dunlay 1987; Hueston 1994; King 1996), one trimethoprim/sulfamethoxazole (Franks 1984), one azithromycin (Evans 2002), one cefuroxime (Matthys 2000), one amoxicillin or erythromycin (Little 2005), two amoxicillin (Little 2013; Nduba 2008) and one co‐amoxiclav (Llor 2013).
The majority of studies used a single reassessment visit to evaluate results of the intervention. The timing of this visit varied from study to study, ranging from two to 14 days after the initiation of treatment. Some investigators also asked patients to keep symptom diaries, which were used to determine the duration of symptoms or disability.
Several of the trials provided results of separate analyses of one or more subsets of patients based on characteristics such as cigarette smoking, patient age, duration of symptoms, presence of purulent sputum or illness severity. All patients enrolled in the study by Nduba 2008 were tested for HIV. We have only included results relating to the subgroup of patients who tested negative. The largest study included in the review, which was incorporated in the current update (Little 2013), was adequately powered for a subgroup analysis of patients aged over 60 years.
For the sensitivity analyses, we included unpublished data from two trials. In one (Howie 1970), patients began self‐treatment with dimethyl chlortetracycline or placebo if a cold or influenza‐like illness was not spontaneously resolving after two days. We included data from a subgroup of patients who had a productive cough prior to beginning treatment. The other study (Kaiser 1996) randomised patients with the common cold to amoxicillin‐clavulanic acid or placebo. We included data from a subgroup who had a concomitant diagnosis of non‐purulent tracheobronchitis, which incorporates 'acute bronchitis'. Further details on the subgroups of patients included from these studies is provided in the Characteristics of included studies table.
Excluded studies
Studies were excluded for a variety of reasons based on study design and intervention criteria. Full descriptions of such exclusions are detailed in the Characteristics of excluded studies table.
Risk of bias in included studies
Sixteen of the 17 included trials were randomised, double or single‐blind evaluations comparing an antibiotic with a placebo. The study added for the 2011 update (Nduba 2008) was the first equivalence RCT included in the review. The earlier study by Little 2005 involved three arms comparing immediate antibiotic therapy, no active treatment or delayed treatment and we included the two arms comparing immediate antibiotic treatment with no treatment only. The study by Llor 2013 added to the current update, involved three arms comparing antibiotic, placebo and anti‐inflammatory treatment; we included data from the antibiotic versus placebo arms. Four reports (Brickfield 1986; Howie 1970; Kaiser 1996; Scherl 1987) did not clearly state the randomisation method used. Only one of the articles (Nduba 2008) reported a formal evaluation of the effectiveness of the blinding procedures used. Compliance or adherence with treatment was measured in six studies; in five, there were no differences in the number of pills taken in the antibiotic and placebo groups (Dunlay 1987; Hueston 1994; Little 2013; Nduba 2008; Stott 1976); in the study by King 1996, 94% of the patients who returned for follow‐up took at least one‐half of their pills and Little 2013 reported > 90% adherence in both groups by day five. Regarding co‐interventions with other medications, four trials asked patients to record the use of non‐prescription medications and included this as an outcome measure (Dunlay 1987; Franks 1984; Hueston 1994; King 1996); one restricted use to aspirin and acetaminophen, but did not have the patients record this (Scherl 1987); and one reported adjunctive prescriptions, but not use of over‐the‐counter medications (Verheij 1994). The majority of studies (13 out of 17) followed up more than 80% of participants (details of drop‐outs are provided in the Characteristics of included studies table). In some cases, no information about withdrawals was available in the paper or from the authors. However, when information was available, we included outcome data from the last point at which the patients were still in the study. As far as possible, we analysed patients on an intention‐to‐treat basis.
The overall risk of bias is presented graphically in Figure 1 and summarised in Figure 2.
'Risk of bias' summary: review authors' judgements about each methodological quality item for each included study.
'Risk of bias' graph: review authors' judgements about each methodological quality item presented as percentages across all included studies.
Allocation
In general, there was minimal risk of allocation or selection bias; 15 out of 17 studies clearly reported adequate allocation concealment.
Blinding
In general, there was minimal risk of bias relating to lack of blinding with 14 out of 17 studies clearly reporting adequate blinding of outcome assessors.
Incomplete outcome data
The majority of studies had adequate completion of outcome data and there was minimal risk of attrition bias.
Selective reporting
Most trials evaluated several different outcome measures. In some cases, the published reports included detailed data for only those outcomes found to be statistically significant. To minimise this reporting bias, we attempted to obtain additional data from the trial authors; five authors provided this information (Howie 1970; Hueston 1994; Kaiser 1996; King 1996; Williamson 1984). However, we were still unable to include data from Stott 1976 for the outcomes of cough, night cough or activity limitations at follow‐up, which were reported in the published trial as being not significantly different between groups.
Other potential sources of bias
The main concern regarding bias was the relatively small numbers of studies that could be included in individual meta‐analyses. We have attempted to address this by adding a new broader analysis reflecting clinical improvement. This has been further strengthened by the addition of the largest multi‐country trial to date (Little 2013). There were no additional concerns regarding other potential sources of bias.
Effects of interventions
The same outcome measures were not reported in all studies. Some studies reported the presence or absence of various symptoms and signs at a follow‐up visit; others reported the mean duration of symptoms; and still others reported only unique symptom scores. Also, in some studies explicit data were available only for outcomes that were significantly different between the antibiotic and placebo groups. Therefore, the number of studies that provide data for the outcomes in this review ranged from three to 11. None of the summary outcomes in the primary analysis exhibited statistically significant heterogeneity apart from the analysis of patient 'clinically improved'. Numbers of studies and participants included in the individual meta‐analyses are generally small, though the meta‐analysis for 'clinically improved' includes 11 studies and the meta‐analysis for adverse events includes 12 studies.
Primary outcomes
1. Cough‐related outcomes
At the follow‐up visit, patients given antibiotics were less likely to have a cough (4 studies with 275 participants, risk ratio (RR) 0.64, 95% CI 0.49 to 0.85, NNT 6) (Analysis 1.1; Figure 3) or have a night cough (4 studies with 538 participants, RR 0.67, 95% CI 0.54 to 0.83, NNT 7) (Analysis 2.1). The differences in presence of a productive cough at follow‐up and days of productive cough did not reach statistical significance. Antibiotic‐treated patients only had a significant reduction in mean duration of cough when the study by Little 2005, which had a no treatment comparison group, was excluded (Figure 4). Llor 2013 also reported no significant difference in the median days of cough between the antibiotic and placebo group. Sensitivity analysis also altered the outcome for the mean duration of productive cough, which was significantly reduced if the Howie 1970 study relating to upper respiratory tract infection was excluded.
Forest plot of comparison: Cough at follow‐up visit, outcome: number of patients with cough.
Forest plot of comparison: 8 Days of cough, outcome: mean number of days of cough.
2. Global assessment of improvement by clinicians at follow‐up: 'clinically improved'
For the 2011 update of the review, we included an additional analysis that included a broader outcome 'clinically improved', so that as many studies as possible could be included in a meta‐analysis. This was particularly important following the inclusion of the Nduba 2008 study in 2011, which was of high quality, included a large number of participants and showed no benefit from antibiotic use. This has been updated and includes additional data from the authors of the largest included study, added to this 2013 update (Little 2013). The data from Little 2013 is based on numbers of patients no longer reporting their symptoms being 'moderately bad' at one week. The published study presents mean symptom severity scores in the first few days, which indicated no significant difference between the intervention and control groups (Little 2013). This outcome reflects the proportions of patients with clinical improvement and incorporates 'cure' as measured by a > 75% reduction in the Acute Bronchitis Severity Score (Nduba 2008), global improvement or being well (Brickfield 1986; Llor 2013; Matthys 2000; Stott 1976; Verheij 1994; Williamson 1984), patient report of no limitations (Dunlay 1987; Evans 2002; Franks 1984) and resolution of symptoms rated as moderately bad, severe or worsening (Little 2013). This analysis includes 11 studies and 3841 participants and shows no statistically significant difference (RR 1.07, 95% CI 0.99 to 1.15, NNT 22) (Analysis 6.1; Figure 5). This is essentially unchanged since the last version of the review although the addition of the data from Little 2013 and Llor 2013 has increased the heterogeneity. A sensitivity analysis removing the studies reporting 'no limitation' made no difference to this result.
Forest plot of comparison: Clinically improved, outcome: number of patients reporting no limitations or described as cured/well/symptoms resolved or globally improved.
3. General clinical outcomes
Antibiotic‐treated patients also had a reduction in the number of days feeling ill (5 studies with 809 participants, mean difference (MD) ‐0.64, 95% CI ‐1.16 to ‐0.13) (Analysis 8.1; Figure 6) and a reduction in days with impaired activity (6 studies with 767 participants, MD ‐0.49, 95% CI ‐0.94 to ‐0.04) (Analysis 9.1). There was no significant difference in proportions of patients with activity limitations at follow‐up. Patients on antibiotics were more likely to be improved by the clinician's global assessment (6 studies with 891 participants, RR 0.61, 95% CI 0.48 to 0.79, NNT 11) (Analysis 10.1; Figure 7) and less likely to have an abnormal lung exam (5 studies with 613 participants, RR 0.54, 95% CI 0.41 to 0.70, NNT 6) (Analysis 11.1). Additional clinical outcomes were reported by Little 2013 who found no significant difference in mean symptom severity scores on days two to four (intervention score 1.62 (standard deviation (SD) 0.84) versus control score 1.69 (SD 0.84), P = 0.07) and Evans 2002 found that azithromycin had no benefit in terms of health‐related quality of life at day three and day seven follow‐up. Llor 2013 also reported no difference in time to overall symptom resolution between groups.
Forest plot of comparison: Days of feeling ill, outcome: mean number of days of feeling ill.
Forest plot of comparison: Not improved by physician's global assessment at follow‐up visit, outcome: number of patients not improved.
Secondary outcomes
1. Adverse effects
With four exceptions (Brickfield 1986; Little 2005; Matthys 2000; Nduba 2008), all of the studies found that patients in the antibiotic group reported more adverse effects than did patients receiving a placebo (Figure 8). The RR of adverse effects in the antibiotic‐treated group was statistically significant at 1.20 (12 studies with 3496 participants, 95% CI 1.05 to 1.36, NNT 24) (Analysis 12.1). The most commonly reported side effects involved gastrointestinal symptoms such as nausea, vomiting or diarrhea. Headaches, skin rash and vaginitis also occurred. Side effects seemed mild as only 0% to 13% (overall 3.7%) of volunteers withdrew for this reason and no deaths were reported.
Forest plot of comparison: Number of patients with adverse effects.
Subgroups
We were not able to obtain enough explicit data from the studies for various patient subgroups, therefore we did not carry out any sensitivity analyses based on patient characteristics (such as age, duration of illness or smoking status). Little 2013 was adequately powered to assess the effect in the subgroup of patients aged over 60 and found no significant benefit in this group. The results in the individual studies for subgroup analyses were mixed. In one trial, all of the significantly improved outcomes from antibiotics occurred in non‐smokers (Brickfield 1986). The other seven trials reported that they found no differences in antibiotic effectiveness for smokers versus non‐smokers but included no data on these comparisons in their published reports. Verheij 1994, using multiple regression, found that two subsets of patients were more likely to improve with doxycycline than placebo: patients over 55 years and patients with very frequent cough who felt ill. Scherl 1987 found that only patients without coryza or sore throat had fewer days of cough or sputum with doxycycline. The only study to use Gram stains (Franks 1984) reported an earlier return to work for patients with a positive Gram stain who were treated with antibiotics. Nduba 2008 also examined whether use of amoxicillin was more effective than placebo in patients who had tested positive for HIV and found no difference, though all patients had received a chest X‐ray and those with any abnormal signs were excluded.
Little 2005 was added to the 2009 update and found no significant difference in outcomes between groups treated with immediate antibiotics compared with no antibiotic treatment. As this study did not involve a placebo control we included it in the analyses, where appropriate data were available, as a subgroup to highlight this difference. The one study included in the 2011 update (Nduba 2008) was powered to detect equivalence between antibiotic and placebo and found no significant difference. In fact, the point estimates favoured placebo treatment (84% cured on placebo versus 82.4%% cured on amoxicillin). The largest included study, which was added in the 2013 update, was included in the meta‐analyses of 'clinically improved' and adverse effects (Little 2013).
Discussion
Summary of main results
There are mixed results across studies with some suggesting marginal benefits for antibiotics, though these are of doubtful clinical significance. However, the inclusion of the largest multi‐centre study of the effectiveness of antibiotics in patients with lower respiratory tract infections strengthens the evidence and also highlights a statistically significant increase in adverse events in the antibiotic‐treated groups. However, it is possible that older patients with multimorbidities may not have been recruited to trials so the evidence guiding decision making in this group of patients is less certain.
Overall completeness and applicability of evidence
In general, the available evidence suggests we should not be using antibiotics to treat acute bronchitis or lower respiratory tract infections when pneumonia is not expected. There is a modest benefit from antibiotics for some outcomes but these are of minimal clinical significance. Any benefit is even less apparent in the sensitivity analysis, which included data from subgroups of patients with productive cough of short duration (two to four days) in conjunction with the common cold. Of the two trials in the primary analysis that limited enrolment to patients who had been ill for less than one week, one did not show any benefit from antibiotics (Stott 1976), whilst the other showed modest benefit with antibiotic (Matthys 2000).
It is possible that the overall benefit noted from antibiotics resulted from the inclusion in some trials of patients who may have had pneumonia instead of acute bronchitis. There was variation between studies on whether chest X‐rays were conducted as part of evaluations. Only one trial (Brickfield 1986) obtained chest radiographs on all patients and then excluded those whose films were consistent with pneumonia. In Little 2013, a positive chest X‐ray was not an automatic exclusion criteria though some patients dropped out following such a finding and further publications on this are planned (author communication). All other studies either excluded or obtained chest radiographs in patients with clinical findings of suspected pneumonia (which in most studies were focal findings on chest examination). Individual signs (such as crackles or fever) are not sensitive (Metlay 1997a), therefore their absence cannot be relied on to rule out pneumonia. On the other hand, since the prevalence of pneumonia in outpatients who present with cough is generally low (less than 5% in the USA) (Metlay 1997b), it is unlikely that a significant number of patients in these trials had pneumonia. Further evidence for this is that only three patients who were randomised to placebo among all nine primary trials were subsequently diagnosed with pneumonia (two patients in Stott 1976 and one in Scherl 1987). In addition, this review is designed to test the effectiveness of treatment for acute bronchitis in clinical practice and it is not standard practice to confirm the diagnosis of acute bronchitis with a chest X‐ray unless there is a clinical suspicion of underlying pneumonia. If we had only included studies with chest X‐ray confirmation of diagnosis it would have limited the generalisability of the review findings.
Quality of the evidence
Since there is no gold standard test, the diagnosis of acute bronchitis must be made on clinical grounds. All of the trials excluded patients with chronic pulmonary disease and enrolled patients with recent onset of a respiratory illness with a productive cough. The results of the studies in the primary analysis that included patients with a productive cough, without specifically stating that the patients had acute bronchitis, were similar to the studies that used this specific terminology, as one showed some benefits from antibiotics (Verheij 1994) and one did not (Stott 1976). Clinical characteristics of patients did vary somewhat among studies regarding the duration of illness and associated symptoms and physical findings but were consistent with definitions generally used by primary care physicians (Oeffinger 1997; Verheij 1990). Therefore, these results would appear to be generalisable to the management of acute bronchitis in community practices.
Potential biases in the review process
This review may also be subject to bias because although we have now included 16 trials and 3656 participants, it is possible that some patient subgroups are under‐represented as they may not have been recruited into the original trials. Little 2013 points out that while they included a large sample of older patients, more severely ill older people with multimorbidities were unlikely to have been approached to participate in the trial and in these types of patients, their results should be interpreted with caution and this applies to the review results also.
Agreements and disagreements with other studies or reviews
In the current update of the review, we have included a large multi‐country trial that shows no benefits from antibiotics even in older patients. Further analyses of the data from this study are ongoing as part of Workpackage 10 of the GRACE program (http://www.grace‐lrti.org). It should be noted that a recent large observational study examining symptom resolution in 2714 patients with acute cough who had been prescribed amoxicillin across 13 European countries found that symptom resolution was quicker in those receiving no antibiotic (Butler 2010).
'Risk of bias' summary: review authors' judgements about each methodological quality item for each included study.
'Risk of bias' graph: review authors' judgements about each methodological quality item presented as percentages across all included studies.
Forest plot of comparison: Cough at follow‐up visit, outcome: number of patients with cough.
Forest plot of comparison: 8 Days of cough, outcome: mean number of days of cough.
Forest plot of comparison: Clinically improved, outcome: number of patients reporting no limitations or described as cured/well/symptoms resolved or globally improved.
Forest plot of comparison: Days of feeling ill, outcome: mean number of days of feeling ill.
Forest plot of comparison: Not improved by physician's global assessment at follow‐up visit, outcome: number of patients not improved.
Forest plot of comparison: Number of patients with adverse effects.
Comparison 1 Cough at follow‐up visit, Outcome 1 Number of patients with cough.
Comparison 2 Night cough at follow‐up visit, Outcome 1 Number of patients with night cough.
Comparison 3 Productive cough at follow‐up visit, Outcome 1 Number of patients with productive cough.
Comparison 4 Days of cough, Outcome 1 Mean number of days of cough.
Comparison 5 Days of productive cough, Outcome 1 Mean number of days of productive cough.
Comparison 6 Clinically improved, Outcome 1 Number of patients reporting no activity limitations or described as cured/globally improved.
Comparison 7 Limitation in work or activities at follow‐up visit, Outcome 1 Number of patients with limitations.
Comparison 8 Days of feeling ill, Outcome 1 Mean number of days of feeling ill.
Comparison 9 Days of impaired activities, Outcome 1 Mean number of days of impaired activities.
Comparison 10 Not improved by physician's global assessment at follow‐up visit, Outcome 1 Number of patients not improved.
Comparison 11 Abnormal lung exam at follow‐up visit, Outcome 1 Number of patients with abnormal lung exams.
Comparison 12 Adverse effects, Outcome 1 Number of patients with adverse effects.
| Outcome or subgroup title | No. of studies | No. of participants | Statistical method | Effect size |
| 1 Number of patients with cough Show forest plot | 4 | 275 | Risk Ratio (M‐H, Fixed, 95% CI) | 0.64 [0.49, 0.85] |
| Outcome or subgroup title | No. of studies | No. of participants | Statistical method | Effect size |
| 1 Number of patients with night cough Show forest plot | 4 | 538 | Risk Ratio (M‐H, Fixed, 95% CI) | 0.67 [0.54, 0.83] |
| Outcome or subgroup title | No. of studies | No. of participants | Statistical method | Effect size |
| 1 Number of patients with productive cough Show forest plot | 7 | 713 | Risk Ratio (M‐H, Fixed, 95% CI) | 0.97 [0.82, 1.16] |
| 1.1 Acute bronchitis studies | 6 | 549 | Risk Ratio (M‐H, Fixed, 95% CI) | 0.88 [0.72, 1.08] |
| 1.2 Subgroup with productive cough from URTI study | 1 | 164 | Risk Ratio (M‐H, Fixed, 95% CI) | 1.24 [0.88, 1.75] |
| Outcome or subgroup title | No. of studies | No. of participants | Statistical method | Effect size |
| 1 Mean number of days of cough Show forest plot | 7 | 2776 | Mean Difference (IV, Fixed, 95% CI) | ‐0.46 [‐0.87, ‐0.04] |
| 1.1 Acute bronchitis studies | 6 | 2350 | Mean Difference (IV, Fixed, 95% CI) | ‐0.55 [1.00, ‐0.10] |
| 1.2 Subgroup with no placebo control | 1 | 426 | Mean Difference (IV, Fixed, 95% CI) | 0.11 [‐1.01, 1.23] |
| Outcome or subgroup title | No. of studies | No. of participants | Statistical method | Effect size |
| 1 Mean number of days of productive cough Show forest plot | 6 | 699 | Mean Difference (IV, Fixed, 95% CI) | ‐0.43 [‐0.93, 0.07] |
| 1.1 Acute bronchitis studies | 5 | 535 | Mean Difference (IV, Fixed, 95% CI) | ‐0.52 [‐1.03, ‐0.01] |
| 1.2 Subgroup with productive cough from URTI study | 1 | 164 | Mean Difference (IV, Fixed, 95% CI) | 1.04 [‐1.04, 3.12] |
| Outcome or subgroup title | No. of studies | No. of participants | Statistical method | Effect size |
| 1 Number of patients reporting no activity limitations or described as cured/globally improved Show forest plot | 11 | 3841 | Risk Ratio (M‐H, Random, 95% CI) | 1.07 [0.99, 1.15] |
| Outcome or subgroup title | No. of studies | No. of participants | Statistical method | Effect size |
| 1 Number of patients with limitations Show forest plot | 5 | 478 | Risk Ratio (M‐H, Fixed, 95% CI) | 0.75 [0.46, 1.22] |
| Outcome or subgroup title | No. of studies | No. of participants | Statistical method | Effect size |
| 1 Mean number of days of feeling ill Show forest plot | 5 | 809 | Mean Difference (IV, Fixed, 95% CI) | ‐0.64 [‐1.16, ‐0.13] |
| 1.1 Acute bronchitis studies | 4 | 435 | Mean Difference (IV, Fixed, 95% CI) | ‐0.58 [‐1.16, ‐0.00] |
| 1.2 Subgroup with no placebo control | 1 | 374 | Mean Difference (IV, Fixed, 95% CI) | ‐0.86 [‐1.97, 0.25] |
| Outcome or subgroup title | No. of studies | No. of participants | Statistical method | Effect size |
| 1 Mean number of days of impaired activities Show forest plot | 6 | 767 | Mean Difference (IV, Fixed, 95% CI) | ‐0.49 [‐0.94, ‐0.04] |
| 1.1 Acute bronchitis studies | 5 | 393 | Mean Difference (IV, Fixed, 95% CI) | ‐0.48 [‐0.96, 0.01] |
| 1.2 Subgroup with no placebo control | 1 | 374 | Mean Difference (IV, Fixed, 95% CI) | ‐0.57 [‐1.75, 0.61] |
| Outcome or subgroup title | No. of studies | No. of participants | Statistical method | Effect size |
| 1 Number of patients not improved Show forest plot | 6 | 891 | Risk Ratio (M‐H, Fixed, 95% CI) | 0.61 [0.48, 0.79] |
| 1.1 Acute bronchitis studies | 5 | 816 | Risk Ratio (M‐H, Fixed, 95% CI) | 0.44 [0.30, 0.65] |
| 1.2 Subgroup with non‐purulent tracheobronchitis from URTI study | 1 | 75 | Risk Ratio (M‐H, Fixed, 95% CI) | 1.03 [0.82, 1.29] |
| Outcome or subgroup title | No. of studies | No. of participants | Statistical method | Effect size |
| 1 Number of patients with abnormal lung exams Show forest plot | 5 | 613 | Risk Ratio (M‐H, Fixed, 95% CI) | 0.54 [0.41, 0.70] |
| Outcome or subgroup title | No. of studies | No. of participants | Statistical method | Effect size |
| 1 Number of patients with adverse effects Show forest plot | 12 | 3496 | Risk Ratio (M‐H, Fixed, 95% CI) | 1.20 [1.05, 1.36] |
| 1.1 Acute bronchitis studies | 11 | 3162 | Risk Ratio (M‐H, Fixed, 95% CI) | 1.22 [1.07, 1.40] |
| 1.2 Subgroup with no placebo control | 1 | 334 | Risk Ratio (M‐H, Fixed, 95% CI) | 0.95 [0.61, 1.50] |
